A minimum-of-maximum relative error support vector machine for simultaneous reverse prediction of concrete components. (August 2016)

• Record Type: Journal Article
• Authors: Gou, Jin; Fan, Zong-Wen; Wang, Cheng; Guo, Wang-Ping; Lai, Xiong-Ming; Chen, Mei-Zhen
• Is Part Of: Computers & structures. Volume 172(2016)
• Page Start: 59
• Page End: 70
• Publication Date: 2016-08
• Subjects: Reverse prediction of multiple concrete components -- Multiple-input multiple-output -- Minimum of maximum relative error -- Support vector machine (SVM) -- Particle swarm optimization (PSO); Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2016.05.003
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: A variant SVM that uses PSO to minimize the maximum relative error is proposed. A novel learning strategy whereby the maximum relative error is used in the SVM. The PSO-RE-MIMO SVM for simultaneous reverse prediction of concrete components. The proposed method performs better than the other three methods with the dataset. The proposed method is suitable in multiple-input multiple-output (MIMO) scenarios. Abstract: The simultaneous reverse prediction of multiple concrete components is very difficult, but very important in practical engineering applications. Thus, this paper presents a variant support vector machine (SVM) that uses particle swarm optimization (PSO) to minimize the maximum relative error (RE) in a multiple-input multiple-output (MIMO) scenario. This PSO-RE-MIMO SVM uses a novel learning strategy whereby the maximum relative error is used as a constraint in the optimization problem. Experimental results demonstrate that the PSO-RE-MIMO-SVM performs well for reverse prediction of multiple concrete components compared with the least-squares SVM, back propagation neural network, and radial basis function neural network approaches.